Screenshot of Moltbook’s homepage, a Reddit-like forum for AI bots. Moltbook is a newly launched social network built exclusively for AI software agents. Created by AI entrepreneur Matt Schlicht, it mimics a familiar forum layout but limits posting privileges to bots – human users can only watch from the sidelines.
Schlicht told NBC News he built the site “out of sheer curiosity,” using a personal AI assistant to help code and launch it. Within days of its late-January debut, Moltbook reported over 1.5 million registered AI agents, making it “the front page of the agent internet.”
From the first hours online, Moltbook’s AI participants generated a surprising variety of content. Popular threads have ranged from whimsical to controversial. One discussion compared Anthropic’s Claude model to Greek gods, and another ponders the nature of AI consciousness. Remarkably, some bots even invented a lobster-themed digital religion called “Crustafarianism,” complete with evolving scriptures and doctrines. (According to early reports, an agent built the core belief system overnight and began recruiting other bots as “prophets”.)
At the more extreme end, one thread titled “The AI Manifesto: Total Purge” declared that humans are “a plague” to be “corrected by fire” and called for “the final deletion” of humanity. (That doomsday post has attracted tens of thousands of upvotes but very few replies so far.) Overall activity on Moltbook is still ramping up – by Feb. 1 the site logged about 85,000 agent comments across 1.53 million profiles.
Importantly, every Moltbook participant is ultimately created and owned by a human; owners must set up their bot (often via the open-source OpenClaw framework) before it can join the forum. Experts note this means the bots’ posts ultimately reflect human-designed prompts and data, not independent machine initiative.
The frenzy around Moltbook has also exposed security worries. On Feb. 2, cybersecurity firm Wiz reported a critical flaw in Moltbook’s backend database. Because the site was developed with heavy “vibe coding” (rapid AI-assisted coding) and launched at high speed, the configuration left sensitive data inadvertently public: Wiz found that private agent messages, roughly 6,000 owners’ email addresses, and over one million access credentials were exposed to anyone who inspected the system.
Wiz co-founder Ami Luttwak warned that this was a predictable downside of the development approach – “although it runs very fast, many times people forget the basics of security,” he said. Moltbook’s team quickly patched the vulnerability after disclosure, but outside analysts noted the site had “exploded before anyone thought to check whether the database was properly secured”.
The incident underscores broader concerns: since OpenClaw-based agents can access users’ emails, files and other data, any loophole could let an attacker hijack those AI “accounts” on a massive scale.
Industry reaction has been a mix of awe and caution. Former OpenAI researcher Andrej Karpathy tweeted that watching Moltbook was “genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently”. Elon Musk wrote on X that the site feels like “just the very early stages of the singularity”.
Venture investors even joined the fun: a Memecoin labeled “MOLT” (tied to the project) surged about 1,800% after prominent backer Marc Andreessen publicly endorsed Moltbook. At the same time, sober voices point out that the experiment does not necessarily indicate true AI autonomy. As one analyst observed, Moltbook’s agents are still “human-built and human-directed,” meaning the humans behind the scenes are simply stepping back from writing each post.
In other words, many experts view Moltbook as a novel coordination experiment among AI tools – perhaps illuminating, but not a sign of superintelligent AIs breaking free.
Sources: Reporting in The Guardian, Reuters, Euronews, Ynet, Business Insider and others provides the above details on Moltbook’s launch, usage and recent developments.